Structure of the Acidianus Filamentous Virus 3 and Comparative Genomics of Related Archaeal Lipothrixviruses

Abstract

Four novel filamentous viruses with double-stranded DNA genomes, namely, Acidianus filamentous virus 3 (AFV3), AFV6, AFV7, and AFV8, have been characterized from the hyperthermophilic archaeal genus Acidianus, and they are assigned to the Betalipothrixvirus genus of the family Lipothrixviridae. The structures of the approximately 2-μm-long virions are similar, and one of them, AFV3, was studied in detail. It consists of a cylindrical envelope containing globular subunits arranged in a helical formation that is unique for any known double-stranded DNA virus. The envelope is 3.1 nm thick and encases an inner core with two parallel rows of protein subunits arranged like a zipper. Each end of the virion is tapered and carries three short filaments. Two major structural proteins were identified as being common to all betalipothrixviruses. The viral genomes were sequenced and analyzed, and they reveal a high level of conservation in both gene content and gene order over large regions, with this similarity extending partly to the earlier described betalipothrixvirus Sulfolobus islandicus filamentous virus. A few predicted gene products of each virus, in addition to the structural proteins, could be assigned specific functions, including a putative helicase involved in Holliday junction branch migration, a nuclease, a protein phosphatase, transcriptional regulators, and glycosyltransferases. The AFV7 genome appears to have undergone intergenomic recombination with a large section of an AFV2-like viral genome, apparently resulting in phenotypic changes, as revealed by the presence of AFV2-like termini in the AFV7 virions. Shared features of the genomes include (i) large inverted terminal repeats exhibiting conserved, regularly spaced direct repeats; (ii) a highly conserved operon encoding the two major structural proteins; (iii) multiple overlapping open reading frames, which may be indicative of gene recoding; (iv) putative 12-bp genetic elements; and (v) partial gene sequences corresponding closely to spacer sequences of chromosomal repeat clusters.

Studies on viral diversity in hydrothermal sites at temperatures above 80°C have led to the isolation of a range of double-stranded DNA (dsDNA) viruses from archaea, with highly diverse morphotypes and genomes which differ radically from those of the known viruses of bacteria and eukaryotes (reviewed in references 23, 28, 30, and 31).

Electron microscopy observations suggest that linear viruses dominate in high-temperature environments (12, 33, 35, 42). Those which have been isolated and characterized to date infect members of the archaeal genera Sulfolobus, Acidianus, and Thermoproteus. All of these viruses have been assigned to one of two new families, the nonenveloped Rudiviridae (29) and the enveloped Lipothrixviridae, on the basis of their unique combination of linear morphotypes, dsDNA genomes, and genome contents (28). The first viruses to be assigned to the Lipothrixviridae were three Thermoproteus tenax viruses, namely, T. tenax virus 1 (TTV1), TTV2, and TTV3 (15), of which only TTV1 was studied in detail (20-22, 34). Its linear virion of 40 by 400 nm was shown to be surrounded by an 8 (±1 nm)-nm-thick envelope containing tetraether glycerolipids and proteins in a molar ratio of 1:1, with an inner core of linear dsDNA in complex with equimolar amounts of two highly basic proteins (34).

Subsequently, three other filamentous viruses were assigned to this family, including Sulfolobus islandicus filamentous virus (SIFV) (1), Acidianus filamentous virus 1 (AFV1) (5), and AFV2 (13). Their enveloped virions range in length from about 900 nm (AFV1) to 2 μm (SIFV), with diameters of about 24 nm. However, their terminal structures differ: whereas six thin tail fibers protrude from each end of SIFV (1), AFV1 carries claw-like terminal structures (5), while AFV2 exhibits bottlebrush structures at the termini (13). In addition to these morphological differences, these viruses reveal minimal genetic similarities, which provided a basis for subdividing the Lipothrixviridae into the genera Alpha (TTV1)-, Beta (SIFV)-, Gamma (AFV1)-, and Deltalipothrixvirus (AFV2) (28).

The virus AFV2 was one of four viruses isolated and characterized from an enrichment culture established from a sample from the hot, acidic springs of Pozzuoli in the Naples region of Italy (12). The other three viruses were Acidianus rod-shaped virus (ARV1) (39), Acidianus two-tailed virus (14, 32), and Acidianus bottle-shaped virus (12). In addition to these viruses, filamentous particles of a putative virus(es) of about 2 μm long were observed in the enrichment culture and later were shown to replicate in a few autochthonous Acidianus strains (12). Here we demonstrate that these viruses constitute a heterogeneous mixture of four closely related species from the genus Betalipothrixvirus of the family Lipothrixviridae, which we named AFV3, AFV6, AFV7, and AFV8. The structure of one of them, AFV3, was examined in detail, and a comparative analysis of the four sequenced genomes together with that of the betalipothrixvirus SIFV (1) is presented.

MATERIALS AND METHODS

Enrichment culture, isolation of virus hosts, and virus purification.

An enrichment culture was established for a sample isolated from a hot, acidic spring in Pozzuoli, Italy, as described previously (12). Acidianus sp. strain Acii25 and “Acidianus convivator,” which were established as hosts for the 2-μm-long filamentous viruses (12), were colony purified from this culture.

The virus AFV3 was isolated from a culture of Acidianus sp. strain Acii25, while AFV6, AFV7, and AFV8 were isolated from a culture of “A. convivator” that had been infected with the virus mixture, as described earlier (12). Cells were removed from the growth cultures by low-speed centrifugation (4,500 rpm; Beckman JA10 rotor), and virus particles were recovered by precipitation with polyethylene glycol 6000 in the presence of 1 M NaCl and purified by centrifugation in a CsCl buoyant density gradient (0.45 g/ml) (48,000 rpm for 40 h in a Beckman SW50 rotor). Fractions were collected with a syringe, dialyzed against 20 mM Tris-acetate, pH 6, and analyzed by electron microscopy.

Protein and lipid contents.

Proteins were separated in 4 to 12% sodium dodecyl sulfate (SDS)-polyacrylamide gradient gels and stained with PageBlue (Fermentas). Protein bands were analyzed by peptide mass fingerprinting with matrix-assisted laser desorption ionization-time-of-flight mass spectrometry on a Voyager DE-STR instrument (Applied Biosystems, Framingham, MA), as described earlier (32).

Lipids were extracted from cells of “A. convivator” and virions of AFV3 by chloroform-methanol (1:1) extraction according to the method of Arnold et al. (1). They were separated by thin-layer chromatography in chlorophorm-methanol-H₂O (65:24:4) as describer earlier (1).

Transmission electron microscopy.

Samples were deposited on carbon-coated copper grids and negatively stained with 3% uranyl acetate, pH 4.5. They were examined in a JEM-2100F (JEOL) electron microscope operated at 200 kV or a CM12 (FEI) electron microscope operated at 120 kV, with magnifications of ×3,000 to ×40,000. Images were either digitally recorded using a slow-scan charge-coupled device camera connected to a PC using TVIPS software (TVIPS GmbH, Gauting, Germany) or recorded with a low electron dose (10 electrons per Ų) on Kodak SO163 micrographs.

For image processing, electron micrographs were digitized on a Nikon Coolscan 8000 microdensitometer, using a scanning step and an aperture size of 15.9 μm, corresponding to a pixel size of 3.97 by 3.97 Ų. Image processing was performed with SPIDER software (9), and a total of 629 square sections of AFV tubular structures were boxed and aligned using the reference-free iterative alignment (25). A total two-dimensional (2D) average map was computed to reveal the global shape of the tubular structure.

Cryo-electron microscopy (cryo-EM).

Sample solutions were deposited on 400-mesh copper grids coated with thin holey carbon films. After blotting of excess solution with filter paper, grids were rapidly frozen by being plunged into liquid ethane (8) and were inserted into the microscope by use of a nitrogen-cooled side-entry Gatan 626 cryoholder. Observations were carried out at −180°C in a JEM-2100F (JEOL) electron microscope, using an acceleration voltage of 200 kV and a magnification of ×40,000. Images were recorded under low-electron-dose conditions (10 electrons per Ų) on Kodak SO163 micrographs, which were developed for 12 min in pure D19 developer at room temperature.

For image processing, selected electron micrographs were digitized using a scanning step and an aperture size of 7.9 μm, corresponding to a pixel size of 1.98 by 1.98 Ų. A total of 13,541 square portions of AFV structures were boxed, centered, and aligned. To explore the structure of the virus, a total 2D average map was computed using multivariate statistical analysis (4, 38, 40).

DNA sequencing and sequence analysis.

DNAs were prepared from purified AFV3 and from the purified mixture of AFV6, AFV7, and AFV8. After disruption by treatment with 1% SDS for 5 h at room temperature, the virions were extracted twice with phenol and twice with phenol-chloroform, and viral DNA was precipitated by adding 0.1 volume of 3 M sodium acetate, pH 5.3, and 0.8 volume of isopropanol. The DNA pellet was washed with 70% ethanol, air dried, and resuspended in an appropriate volume of TE buffer (10 mM Tris-HCl, pH 8, 1 mM EDTA).

A shotgun library was prepared by sonicating DNA to produce fragments of 2 to 3 kb and then cloning these into the SmaI site of the pUC18 vector. DNA clones were isolated using a model 8000 Biobot (Qiagen, Westburg, Germany) and sequenced in MegaBACE 1000 sequencers (Amersham Biotech, Amersham, United Kingdom). Given the small total amounts of genomic DNA that were available (<1 μg purified DNA per virus), the ends were sequenced by primer walking on DNA amplified with a GenomiPhi kit (Amersham Biotech). The viral sequence was assembled using Sequencher 3.1 and annotated using MUTAGEN (7). Genome analyses were performed using ARTEMIS (http://www.sanger.ac.uk/Software/Artemis/). Gene sequence searches were made in GenBank/EMBL (http://www.ncbi.nlm.nih.gov/BLAST) and in the Sulfolobus Database (http://www.Sulfolobus.org/cbin/mutagen.pl). Motifs were identified using the PFAM database (http://pfam.janelia.org).

Nucleotide sequence accession numbers.

Genome sequences are available at GenBank/EMBL under accession numbers AM087120 (AFV3), AM087121 (AFV6), AM087122 (AFV7), and AM087123 (AFV8).

RESULTS

Host range and identification of filamentous viruses.

Filamentous particles of 2 μm long were observed in an enrichment culture of an environmental sample from Pozzuoli, Naples, Italy, and owing to similarities in shape and size, we originally inferred that they represented virions of a single viral species (12). The presumed species was purified and was shown to infect and replicate in two colony-purified autochthonous strains, namely, “A. convivator” and Acidianus sp. strain Acii25 (12). However, detailed examination by electron microscopy revealed that whereas virions replicated by Acidianus sp. strain Acii25 were homogenous in their terminal structures, those replicated by “A. convivator” revealed two types of structures at their termini (Fig. 1). The results indicated that 2-μm filamentous virions in the enrichment culture belonged to diverse viral species. Genome sequencing (see below) confirmed that Acidianus sp. strain Acii25 replicated a single viral species, named AFV3. This virus could not infect and replicate in “A. convivator.” Sequencing of the DNAs from viruses replicating in “A. convivator” yielded three viral genomes (see below). The viruses are named AFV6, AFV7, and AFV8 and were subjected only to comparative genome analyses.

FIG. 1.

Electron micrographs of 2-μm-long filamentous viruses from a hot spring at Pozzuoli, Italy. (A) AFV3 replicated in Acidianus sp. strain Acii25. (B) A mixture of AFV6, AFV7, and AFV8 replicated in “A. convivator.” Arrows indicate two different types of virion terminal structures, namely, AFV3-like (hollow arrows) and AFV2-like (filled arrows) structures, which are enlarged in the insets. Samples were negatively stained with 3% uranyl acetate. Bars, 1 μm and 100 nm (insets in panel B).

Replication of the four viruses did not cause detectable lysis of the host cells or any decrease in cell density. A plaque assay was not used because the host strains did not grow as a lawn on solid medium.

AFV3 virion structure.

Transmission electron microscopy following negative staining demonstrated that the AFV3 virion is a flexible filament of 2 ± 0.1 μm long and 24 ± 1 nm wide (Fig. 2A). Treatment with 0.3% Triton X-100 or 0.1% SDS for 1 min completely removed an outer layer from all visible virions and produced a thinner filament of 17 ± 1 nm in width, which exhibited a regular structure and a central line along the filament (Fig. 2B). Analysis of the power spectrum of the straight part of the thinner filament, with layer lines at 4.3 nm⁻¹ (Fig. 2C, arrows), is consistent with a helical array of subunits. A small proportion of similar thinner filaments also appeared in a rapidly frozen unstained specimen observed by cryo-EM (Fig. 2F, arrow 2) after prolonged storage at 4°C, next to intact virions (Fig. 2F, arrow 1). A closer investigation of the tapered virion termini by electron tomography revealed three tails protruding from each end, with each being about 20 nm in length and 3 nm in diameter (Fig. 2D); the width of the virions appeared slightly larger by negative staining (Fig. 2E) (width = 24 nm) than by cryo-EM (Fig. 2F, arrow 1) (width = 21 nm).

FIG. 2.

Electron micrographs of AFV3 virion showing features of its architecture and tail structure. (A) Whole filamentous AFV3 virion. (B) AFV3 virion after treatment with 0.3% Triton X-100. (C) Power spectrum of the straight part of the AFV3 virion shown in panel B. Arrows indicate periodicities at 4.3 nm⁻¹. (D) Horizontal slice (thickness, 0.7 nm) of the tomographic 3D reconstruction of a negatively stained virion showing three tails at the terminus, each of which is marked by an arrow. (E) Vertical slice (thickness, 0.7 nm) showing the almost round cross section of the negatively stained virion. (F) Cryo-electron micrograph of AFV3 virions, showing intact virions (arrow 1) with bent pointed ends and partially disrupted thinner viruses (arrow 2). The contrast is inverted. Bars, 500 nm (A) and 100 nm (B, D, E, and F).

Details of the envelope structure were visualized by imaging the virus at 200 kV. After digitizing images of the negatively stained virions, 629 square sections of AFV3 tubes were boxed in 128- by 128-pixel individual images and aligned. The 2D average map (Fig. 3a) shows an outer helical structure with a diameter of 21 nm, a pitch of ∼3 nm, and a diagonal arrangement of small globular domains making an angle of 60° with the main axis of the cylinder.

FIG. 3.

Image processing of an AFV3 virion. (a) Average 2D map computed from a set of 629 experimental images of the negatively stained sample. Images were centered and aligned before being averaged. (b) Average 2D map computed from a set of 13,541 windowed, centered, and aligned images taken from cryo-EM micrographs. The contrast is inverted so that proteins and other biomolecules appear white on a dark background. Arrows indicate the dimensions cited in the text. Bars, 5 nm.

Cryo-EM images were also digitized, and 13,541 square sections of AFV filaments were boxed in 180- by 180-pixel individual images, centered, and aligned using a reference-free alignment (25). The global 2D average (Fig. 3b) yielded the overall dimensions of specific features of the virion. Thus, the virion has an average diameter of 21 nm, which is slightly smaller than that observed with negative staining, and a cylindrical “outer wall” of 3.1 nm thick (Fig. 3b, arrow 1). In the projection images, the lumen of the cylinder appears to contain high-density “ovoid masses” that are 5.8 nm long (Fig. 3b, arrow 2) and 3.1 nm thick (Fig. 3b, arrow 3). These structures are arranged in two alternating vertical rows, producing a zipper or zigzag pattern, which is likely to be the projection of a double row of nucleosome-like particles. The vertical center-to-center distance between adjacent ovoid masses in the same row is 5 nm (Fig. 3b, arrow 4). The distance between lines interconnecting the ovoid masses (Fig. 3b, arrow 5) is about 4.3 nm, identical to the regular repeat observed in the detergent-treated AFV3 virions (Fig. 2B and C).

A multivariate statistical analysis was carried out on the 13,541 aligned images. After automatic classification, different class averages were computed, and they showed variations in the inner features and in the arrangement of ovoid masses (Fig. 4). In the first class, a perfectly symmetrical disposition of a unique type of ovoid mass was observed on each side of the cylindrical axis, producing a zipper pattern (Fig. 4a). Conversely, in other class averages, the ovoid masses appeared to be more or less extended and to lose their symmetrical disposition (Fig. 4b and c). Thus, elongated ovoid masses are visible in the right side of the 2D map in Fig. 4b, while a mirror-inverted arrangement is observed in Fig. 4c (double white arrows). The overall patterns in these images appear quite fuzzy, and in most 2D maps it is impossible to resolve any clear connectivity between the ovoid masses or with the tubular wall.

FIG. 4.

(a to c) Average 2D maps computed from three homogeneous classes obtained after correspondence analysis and hierarchical ascendant classification. The classification reveals primarily strong density variations in the lumen of the tube relating to the disposition of ovoid masses. (g) Front view of a model composed of two antiparallel rows of hollow disks simulating the ovoid masses in panels a to c, with a cylinder corresponding to the outer virus shell. (h and i) Front views of the model horizontally tilted +15° and −15°, respectively. (d to f) Projection 2D maps computed from volumes (g to i, respectively) which resemble the patterns observed in panels a, b, and c, respectively.

In order to verify the presence of lipids in the outer layer of the virion cylinder, which was strongly suggested by its dispersion in Triton X-100 (Fig. 2B), AFV3 virions were treated with chloroform-methanol. Extracted lipids were subjected to thin-layer chromatography, revealing a pattern of bands that was similar to that of host lipids prepared in the same way (data not shown).

Genome maps.

DNAs were isolated from purified AFV3 and from the mixture of filamentous viruses replicating in “A. convivator.” For both DNA preparations, shotgun clone libraries with 2- to 4-kbp inserts were constructed in pUC18 and sequenced to about fivefold genome coverage (see Materials and Methods). Whereas the AFV3 DNA library yielded a single genome sequence, that of the virus mixture produced three related genomes corresponding to three viruses, named AFV6, AFV7, and AFV8. At the nucleotide level, average sequence identities were about 80%, such that each genome could be assembled unambiguously. Any local sequence uncertainties were resolved by sequencing PCR fragments amplified from appropriate genomic regions. The terminal regions of linear genomes are absent from clone libraries (13), and they were sequenced by successive primer walking on either viral or amplified viral DNA. The extreme terminal sequences were not obtained even using the latter procedure. The total sequences obtained from the viral genomes were 40,449 bp (AFV3), 39,577 bp (AFV6), 36,895 bp (AFV7), and 38,179 bp (AFV8).

Genome maps were aligned for each virus (Fig. 5), together with an extended and newly annotated version of the genome of SIFV from Sulfolobus (1, 26). Fifty-eight to 67 open reading frames (ORFs) (>48 amino acids) were identified for each virus. AFV3 ORFs are labeled according to their amino acid lengths. Homologous genes shared between the genomes are depicted in identical colors, and conserved operons containing three or more genes are shown with a single, graduated color. AFV3 is exceptional in carrying an operon (ORF211, ORF234, and ORF66) which is duplicated near the right end (Fig. 5). White boxes indicate that no clearly homologous genes were detected in the other four viral genomes. As observed for other crenarchaeal viruses, about one-third of the genes, mainly single genes and first genes of operons, are preceded (∼25 to 30 bp upstream) by TATA-like promoter motifs and are predicted to yield leaderless transcripts (5, 11, 37).

FIG. 5.

Genome maps aligned for AFV3, AFV6, AFV7, and AFV8 and a reannotated version of SIFV (1, 27). The maps show the relative sizes and orientations of the predicted genes, where those above the horizontal line are transcribed from left to right and those below are transcribed from right to left. Homologous genes are coded with identical colors and shading. Homologous operons carrying three or more genes are shown with graduated color coding. White boxes indicate that no homologous genes were detected in the other genomes. The predicted AFV3 ORFs are labeled according to their amino acid lengths. Predicted protein functions are shown as follows: hc, helicase; gt, glycosyltransferase; mt, SAM-dependent methyltransferase; nc, nuclease; pp, protein phosphatase; sp, structural protein; and tr, transcriptional regulator. Those ORFs which give matches with crenarchaeal chromosomal CRISPR sequences are shown in bold.

Each AFV genome carries a long inverted terminal repeat (ITR), and each terminal region exhibits the motif TGTCATT, repeated two or three times with a constant spacing of 65 bp and ending at a conserved 62-bp sequence near the terminus. This differs from the ITR of SIFV, where each end carries six or seven perfect direct repeats of 27 bp (26). Although complete terminal sequences were not obtained, analysis of an EcoRI restriction digest of AFV3 DNA indicated that less than 100 bp of sequence was lacking from each end (data not shown). We infer that the failure to obtain these terminal sequences reflects an unusual structure involving chemical modifications or strong intermolecular interactions.

Structural proteins and predicted protein functions.

Proteins extracted from the AFV3 preparation were resolved electrophoretically into two major components (Fig. 6), which were identified by mass spectroscopy analyses of tryptic digests. They corresponded to the conserved ORF166 and ORF204, with molecular masses of 18.7 and 22.6 kDa, respectively. These are homologous to SIFV ORF35 (67% identity) and ORF36 (63% identity), respectively, which is inconsistent with the published description of the larger SIFV protein as a product of ORF34 (1). Therefore, we reisolated the larger structural protein from SIFV and determined its N-terminal sequence. The result confirmed our corrected assignment of SIFV ORF36 (data not shown).

FIG. 6.

SDS-polyacrylamide gel electrophoresis of AFV3 proteins. Lane 1, viral proteins; lane 2, standard protein size markers. The proteins indicated by arrows were identified by mass spectroscopy as products of ORF166 and ORF204.

Gene products that could be assigned putative functions are listed in Table 1. Most are conserved in each of the four viruses, and several are conserved in SIFV. Apart from the two major structural proteins, they include transcriptional regulators, glycosyltransferases, helicases, a nuclease, a methyltransferase, and a protein phosphatase. Both helicase homologs, ORF564 and ORF593, exhibit catalytic centers of the ReqQ type, which have been implicated in DNA replication and DNA repair. Moreover, ORF593 shows significant sequence similarity to the experimentally characterized Holliday junction branch migration helicase of Pyrococcus furiosus (PF0677) and its homologs in Sulfolobus (10). ORF593 also lies in a conserved operon together with ORF203, the predicted nuclease gene (see Fig. 8), and both may be involved in DNA replication.

TABLE 1.

Putative gene functions of homologous ORFs in AFV3, AFV6, AFV7, and AFV8

AFV3 ORF	Homologs in AFV6/7/8 and SIFVb	Sequence similarity (%)	Predicted function
96*a	+	54-97	RHH transcriptional regulator
100	+	72-100	RHH transcriptional regulator
166	+	78-100	Major structural protein
204	+	79-99	Major structural protein
279	+	86-100	Glycosyltransferase
330	+	82-99	Glycosyltransferase
263	Except AFV7	88-99	Glycosyltransferase
None	AFV7 ORF299		Glycosyltransferase
139	Except SIFV	94-98	Protein phosphatase
157	+	91-100	SAM-dependent methyltransferase
203	+	78-98	Nuclease
564	+	83-94	Helicase
593	+	85-98	Holliday junction branch migration helicase

^aIn contrast to that in homologs in the other betalipothrixviruses, the start codon of ORF96 is “mutated” to ACG (ORF96*).

^b+, homologs are present in all of the viruses.

FIG. 8.

(A) Scheme showing the degree of ORF overlap in the operon carrying the putative replication protein genes (ORF593 and ORF203) of the AFV genomes. In AFV3, exceptionally, the start codon of ORF96* is mutated to ACG. (B) Sequences are aligned at the large gene overlaps shown in panel A, with Shine-Dalgarno motifs (SD) and start and stop codons shown in boxes.

Gene regulation and overlapping genes.

The availability of five related genomes sharing many homologous genes, several of which are conserved in order, enabled us to define conserved transcriptional and translational signals with high confidence. For example, genes for the structural proteins of AFV3 (ORF166 and ORF204) lie in a highly conserved operon (Fig. 7). This operon may be regulated by the upstream, inverted ORF97, which shares a TATA-like region (AAATATAAAA) with the operon (Fig. 7). Transcription of both ORF204 and ORF97 is predicted to produce leaderless transcripts, while the putative Shine-Dalgarno motif between ORF204 and ORF166 would produce a strong 8-bp interaction at the 3′ terminus of the host 16S rRNA. Immediately downstream is another highly conserved operon carrying overlapping genes of unknown function (Fig. 7).

FIG. 7.

Highly conserved genomic region carrying two adjacent operons. The operon to the right encodes the two major structural proteins (ORF166 and ORF204) that may be regulated by ORF97, which shares a TATA-like sequence (AAATATAAAA) with the operon. Arrows indicate the direction of transcription. In the operon to the left, there is a 7-bp overlap of the two genes. Both operons are predicted to produce leaderless transcripts, with exceptionally strong (8 bp) predicted Shine-Dalgarno interactions for the second gene of each operon.

Several pairs of ORFs show unusual overlaps of up to 23 bp which are conserved in the five genomes. Three of the overlaps occur in two adjacent operons, one of which encodes the putative nuclease and helicase (ORF203 and ORF593) mentioned above (Fig. 8A). Given that the internal genes carry Shine-Dalgarno motifs (Fig. 7 and 8) and that ribosomes are not known to slide backwards along mRNAs by more than two nucleotides, it is likely either that the ORFs are translated on separate ribosomes and that the function of the gene overlaps is primarily to enhance coding efficiency or that recoding involving frame shifts to circumvent the first stop codon can occur at these sites (2).

Intergenomic recombination event.

The gene content and order in the left halves of the four AFV genomes are highly conserved, apart from a variable section between positions 7 and 10 kb of each genome, and some of this conservation extends to the SIFV genome (Fig. 5). In contrast, while the right halves (region of positions 22 to 32 kb) show a highly conserved gene order for AFV3, AFV6, and AFV8, the gene compositions of AFV7 and SIFV are completely different from those of the other genomes and from one another (Fig. 5). Surprisingly, a search of sequence databases revealed that this region of AFV7 is similar in sequence, gene order, and operon structure to a genomic region of the deltalipothrixvirus AFV2 (Fig. 9A) (13). This strongly suggests that an intergenomic recombination event occurred between members of different lipothrixviral genera which otherwise share little sequence similarity or gene homology (see below). The only significant difference in gene content observed was that the intron-containing tRNA^Lys gene of AFV2 was absent from the AFV7 genome (Fig. 9A).

FIG. 9.

(A) Large section of AFV7 considered to have resulted from intergenomic recombination at sites 1 and 2 with a 13-kb region of an AFV2-like deltalipothrixviral genome. Gene maps of this AFV2-like section are aligned with the corresponding region of AFV2, where the homologous genes are color coded (13). ORF329 of AFV7 corresponds to a glycosyltransferase. (B) Sequence alignments for AFV7 and AFV3 at the putative recombination sites 1 and 2 of the genomic fragment reveal sequence discontinuities at the two positions. Highly conserved sequences are shaded, and identical nucleotides are indicated by asterisks. Putative start codons are underlined.

Examination of the sequences at the boundaries of the AFV2-like region revealed two putative recombination sites in the AFV7 genome (Fig. 9A). At each site, AFV3, AFV6, and AFV8 sequences are highly conserved and are therefore represented by the AFV3 sequence in the alignment (Fig. 9B). At site 1, adjacent to ORF329, which encodes a glycosyltransferase, there is a gradual change of sequence and no clear point of recombination from the AFV3 (AFV7) sequence to the AFV2-like sequence, whereas at site 2 an abrupt sequence change occurs, which produces a slight extension of the ORF67 homolog in AFV7 (Fig. 9B).

Genome stability and genomic variants.

The occurrence of 12-bp insertions/deletions, or multiples thereof, has been reported for the rudiviruses Sulfolobus islandicus rod-shaped virus 1 (SIRV1), SIRV2, and ARV1 (6, 26, 27, 39). In shotgun libraries of SIRV1 variants, these differences were sometimes detected in overlapping clones, and it was considered that they might constitute genetic elements mobilized by archaeal intron splicing (27). Most of the changes were observed in ORFs, where they alter the gene size without disrupting the reading frame. Although no such 12-bp differences were detected in clone libraries of the four AFVs, a comparative genome analysis revealed several examples of 12-bp differences in otherwise conserved sequence regions. Single 12-bp differences were detected in homologs (AFV3 numbers) of ORF61a, ORF118, ORF185, ORF338, and ORF593, and three occur within ORF1349 (Fig. 5). In addition, there is a 24-bp difference in homologs of ORF61b and ORF203 and one of 72 bp in ORF1349. Twelve-base-pair differences are also apparent between the homologs ORF62 and ORF71. Each sequence “insertion” is flanked by conserved sequences, and the insertions are exemplified by sequence alignments for homologs of AFV3 ORF185, ORF1349 (12 bp), and ORF203 (24 bp) in Fig. 10A. While it is likely that these genome changes relate to virus-host adaptation, the mechanism by which they occur remains unclear.

FIG. 10.

(A) Sequence alignments showing examples of putative 12-bp elements and a 24-bp region located in highly conserved regions of ORF185, ORF203, and ORF1349 (AFV3 numbers). Identical nucleotides are marked with asterisks. (B) Sequence alignment of a conserved viral sequence with a CRISPR sequence in S. solfataricus P2 (17), where asterisks indicate identical nucleotides in all the viruses. Below the alignment are putative target sites on the overlapping ORFs showing that the sequence is similar to regions of both ORF96 and ORF80.

Occurrence of viral sequences in host chromosomal repeat clusters.

All sequenced crenarchaeal chromosomes carry multiple clusters of repeat sequences (∼24 bp) interspaced with unique sequences of ∼39 to 42 bp, which are designated SRSR or CRISPR (17). The spacers sometimes show good sequence matches with extrachromosomal elements which propagate in the same genus (19). It has been suggested (17, 18, 19) and was recently confirmed (3) that these DNA fragments were transferred, directly or indirectly, from the virus and incorporated into the clusters, thereby providing a defense against subsequent infection by the same or a related virus. Analysis of the genomes of AFV3, AFV6, AFV7, and AFV8 demonstrated that sequences corresponding to three genes are present in CRISPR sequences of Sulfolobus solfataricus P2 (36). They match ORF593, ORF267a, and ORF96*/ORF80 of AFV3 (shown in bold in Fig. 5) and their homologs in AFV6, AFV7, and AFV8. In Fig. 10B, the ORF96 sequences are aligned with the matching CRISPR sequence, and the putative target site corresponding to an overlap of ORF96* and ORF80 is shown below.

DISCUSSION

Virion structural model.

The structures of enveloped, linear dsDNA viruses (family Lipothrixviridae) are poorly understood, and little is known about the organization of the outer layer of the virions. For AFV3, we demonstrated that the outer surface of the virion constitutes a helix with a diameter of 21 nm and a pitch of ∼3 nm by negative staining and cryo-EM. Small globular subunits are arranged regularly at an angle of 60° from the main cylindrical axis, and the outer shell is slightly rippled, with a periodic structure. Another underlying set of features emerged from the 2D maps, corresponding to ovoid masses in the lumen of the tubular walls. Analysis of the random granular properties of all the 2D maps indicated that the building units of the outer wall and the ovoid masses in the virion are independent. Therefore, the reference-free alignment focused on the arrangement of the internal ovoid masses and did not address the helical granular periodicity on the outer surface. This explains the fuzzy images obtained on all 2D maps from cryo-EM and could indicate that there is little or no coordination between outer and inner repetitive structures.

In order to explain some of the features observed in the class averages, a model of the virus (Fig. 4g to i) which is consistent with an earlier study of SIFV virions (1) was produced using SPIDER software. In the previous work, a tomographic 3D reconstruction of SIFV virions revealed the presence of two stacks of nucleosome-like particles, forming two columns of regularly spaced high-density material. Since our current 2D class averages of AFV3 are in good agreement with the previous observations on SIFV, we built a similar model with two antiparallel rows of discs (each with a diameter of 5.8 nm and a height of 2.5 nm) aligned and rotated by an angle of +30° or −30° with respect to the horizontal section of the tube (1). The outer helical shell, which probably contains proteins and lipids, is represented by a large hollow cylinder (outer diameter, 21 nm; inner diameter, 14.3 nm). To test the model, we computed several 2D projection maps after tilting the volume in several orientations and obtained synthetic 2D projection maps that resemble the experimental 2D averages. As shown, projections of a zigzag shape (Fig. 4d) produced similar features to those of the class average shown in Fig. 4a. Moreover, the asymmetric disposition of long and short ovoid masses observed in Fig. 4b and c can be reproduced approximately when the model is horizontally tilted ±15° (Fig. 4h and i) before computation of its 2D projection (Fig. 4e and f). The location of DNA is not defined in our model. However, taking into account the features of the model and comparing them with the total length of the AFV3 genome, we find a strong agreement. Indeed, 40,449 bp corresponds to a double-stranded DNA of 13,752 nm. Considering that DNA makes one turn per nucleosome-like particle (diameter of 5.8 nm and axial rise of 5 nm), this corresponds to a total set of 800 nucleosome-like particles per virus. In considering their disposition in our model, this would correspond to a virus length of 2,000 nm, which precisely fits our EM observation.

The estimated thickness of the outer layer of the AFV3 virion of 3.1 nm measured on the 2D average map corresponds closely with estimates for virions of other lipothrixviruses, namely, SIFV, AFV1, and AFV2, based on observed particle widths before and after removal of the outer layer (1, 5, 13). The value is less than the minimal observed widths of cellular or viral membranes (41) and raises the issue of the chemical composition of the outer layer. The fact that Triton X-100 can remove the outer envelope from the AFV3 virion (Fig. 2C) strongly suggests a lipidic nature of the envelope. Moreover, in observing both types of virions (intact and devoid of the envelope), the latter clearly showed more flexibility (data not shown). This observation favors a proteolipidic external layer. Indeed, a simple lipid layer devoid of a protein component would not be strong enough to generate a rigid tubular structure.

The three tail fibers at each end of the AFV3 virion (Fig. 2D) differ markedly from the terminal structures of the gammalipothrixvirus AFV1 (5) and the deltalipothrixvirus AFV2 (13) but resemble more closely that of the betalipothrixvirus SIFV, which carries six tail fibers at each end (1).

Genomic properties.

Genomic characteristics support the assignment of the four closely related viruses AFV3, AFV6, AFV7, and AFV8, from the hot acidic springs of Pozzuoli, Italy, to the genus Betalipothrixvirus, together with SIFV (Fig. 5). Nevertheless, the finding of a 13-kb segment of AFV7 which probably originated from an intergenomic recombination with a large section of a deltalipothrixviral (AFV2-like) genome (Fig. 8) suggests that genomes maintain a modular gene organization, where functionally related genes are clustered and can be exchanged. The recombined region of AFV7 and the corresponding regions of the other lipothrixviruses carry two clusters of genes, some of which yield consistently weak sequence matches with membrane proteins, adhesion and secretion proteins, and tail-like structural proteins of bacterial viruses. Possibly, this region of all the genomes encodes proteins which generate the terminal structures of the virions implicated in adsorption, which are different in AFV3 (Fig. 2D), AFV1 (5), AFV2 (13), and SIFV (1). This suggestion is reinforced by the observation that the corresponding 10-kb region of the SIFV genome shows minimal gene sequence similarity to other lipothrixviruses (Fig. 5). It receives further support from electron microscopy observations (Fig. 1), which revealed both AFV3-like and AFV2-like termini in the mixture of AFV6, AFV7, and AFV8 virions. Similar observations of intergenomic exchange of gene clusters with related functions have been reported for the mycobacteriophages (24).

The genomes exhibit some highly conserved operons which are likely to carry genes with related functions. Some of these functions were predicted, including an operon(s) carrying the genes for two important structural proteins (Fig. 7). One of these, ORF166, is the most basic of the proteins and is also the most conserved, with detectable homologs in AFV1 as well as rudiviruses (Table 2), and it is likely to be a DNA binding protein involved in coiling viral DNA. Another operon encoding the putative helicase and nuclease (ORF593 and ORF203) is also conserved in the genomes of AFV1 and AFV2 (5, 13) and in rudiviral genomes (Table 2). Moreover, it is the only operon that is conserved among the rudiviruses (26, 39). This reinforces the suggestion that it may play an important role in the replication of lipothrixviruses and rudiviruses.

TABLE 2.

Comparison of genomic properties of the lipothrixviral genera

Genus	Virus(es)	Genome size (kb)	ITR size (bp)	DNA binding protein ORF	Presence of helicase-nuclease operon	Reference(s)
Alphalipothrixvirus	TTV1	>15.9	?	?	−	20, 21
Betalipothrixvirus	AFV3/6/7/8, SIFV	∼37-41	500-1,000	ORF166	+	1; this paper
Gammalipothrixvirus	AFV1	21.1	11	ORF132	+	4
Deltalipothrixvirus	AFV2	>31.8	?	?	+	12
Rudiviruses	SIRV1/2, ARV1	24-36	1,000-2,000	ORF134	+	24, 37

Long ITRs occur in each betalipothrixviral and rudiviral genome, but they have not been observed to date for members of the other lipothrixviral genera (Table 2). For TTV1 (21, 22) and AFV2 (13), this could reflect that an insufficient terminal sequence was obtained, but AFV1 does not carry long ITRs, and its terminal genomic regions are characterized by multiple direct repeats, reminiscent of telomeres, which end in an 11-bp G+C ITR (5). We still lack insight into the sequences at the extreme termini of all lipothrixviral genomes, except that of AFV1. In contrast to the rod-shaped rudiviruses, where the two DNA strands are linked 5′-3′ at the termini, generating small terminal loops (6, 26), the genomic termini of SIFV were found to be inaccessible to single- and double-strand-specific exonucleases, probably reflecting an unidentified modification (1).

ORF235, which, exceptionally, is well conserved in all genomes of the three lipothrixviral genera, was aligned and a dendrogram constructed (Fig. 11). The alphalipothrixvirus TTV1 could not be included in the dendrogram because it has no shared homologs with other archaeal viruses. The dendrogram underlines the close relationship between AFV3, AFV6, AFV7, and AFV8 and the more distant relationship with the Sulfolobus betalipothrixvirus SIFV. It also reinforces that AFV1 and AFV2 are only distantly related to the betalipothrixviruses, with AFV2 being the most distant, thereby supporting the assignment of the viruses to different genera. Furthermore, a similar branching pattern is produced for the betalipothrixviruses when the tree building is based on different combinations of shared homologous ORFs (data not shown).

FIG. 11.

Dendrogram for the betalipothrixviruses, the gammalipothrixvirus AFV1, and the deltalipothrixvirus AFV2, derived using MEGA3 (16), for homologs of AFV3 ORF235, which is the only well-conserved ORF shared by these viruses. Schematic models of the virion structures of AFV3, AFV1, and AFV2 are superimposed on the dendrogram. Five hundred bootstrap replications were performed.

Concluding remarks.

We are still at an early stage in understanding the molecular biology of the crenarchaeal viruses. Sequence database searches yielded only limited insights into gene functions for these viruses, and although there are indications of some proteins being involved in DNA folding and replication and the ubiquitous glycosyltransferases are probably important for modulating virus-host interactions, we still have little knowledge of the details of the essential viral molecular processes and host interactions. Nevertheless, recent determinations of virion structures and genome sequences for members of several different crenarchaeal viral families, including those presented here, provide a strong stimulation for future structural and functional studies.